Skip to main content
Premium Trial:

Request an Annual Quote

Pan-Cancer Genomic Data Analysis Identifies Prognostic Biomarkers

NEW YORK – Findings from a large-scale pan-cancer genomic data analysis suggest that copy number and gene expression features found in tumor genomes may be more frequently linked to patient outcomes than point mutations.

"To improve our ability to identify the most aggressive malignancies, we constructed genome-wide survival models using gene expression, copy number, methylation, and mutation data from 10,884 patients," senior author Jason Sheltzer, a researcher at Yale University School of Medicine, and coauthor Joan Smith, an investigator affiliated with Yale and Google, wrote in a Cell Reports paper published on Tuesday.

Using data generated for tumors from 33 cancer types for the Cancer Genome Atlas project, combined with clinical outcomes, the researchers searched for point mutations, copy number alterations, gene expression profiles, microRNAs, DNA methylation features, or protein expression patterns showing ties to patient outcomes.

Based on overall survival data for two dozen cancer types and progression-free survival outcomes in nine more types of cancer, they tracked down more than 100,000 biomarkers with significant survival associations.

By digging into the proposed prognostic biomarker set, the two researchers found that certain types of genomic features were more apt to correspond with survival outcomes than others, including copy number changes, transcriptomic features, and epigenetic marks. In contrast, point mutations appeared less likely to track with patient outcomes.

"In general, gene expression, DNA methylation, and [copy number alterations] provided the most prognostic information, while mutational analysis provided the least," the authors reported, adding that "[c]ancers that arose from related tissues of origin tended to display similar survival profiles."

Likewise, prognostic markers were only rarely linked to alterations affecting cancer-promoting oncogenes or targets for existing drug treatments, the researchers reported, despite predicted relationships between cancer drivers and prognostic outcomes.

Rather, the investigators' gene enrichment analyses uncovered relationships between patient outcomes and enhanced expression of cell cycle- and proliferation-related genes, along with other essential housekeeping genes.

"While adverse biomarkers are commonly believed to represent cancer driver genes and promising therapeutic targets, we show that cancer features associated with shorter survival times are not enriched for either oncogenes or for successful drug targets," the authors wrote.

As reported in the past, the findings highlighted the importance of established prognostic factors such as cancer stage, tumor grade, and an individual's age at diagnosis, which were independently associated with patient outcomes across cancer types. The team saw similar overlap for certain biomarkers with pan-cancer links to patient outcomes, though the analysis did not point to specific features that were prognostic in all of the cancers considered.

"Our study illustrates the unexpected prognostic potential of different classes of genomic data," the authors reported, adding that the results "also have significant implications for the analysis of cancer survival data in a preclinical or therapeutic discovery setting."